Kanwas: Open-source shared context board for teams and AI agents
Kanwas is an open-source shared context board that lets teams and AI agents work on the same canvas simultaneously. Agent tool calls stream into a shared timeline visible to everyone. The project is Apache 2.0 licensed and available on GitHub.
Key features
- Multiplayer workspace: Teams and agents share documents, evidence, and decisions on one canvas.
- Live agent streaming: AI agent tool calls appear in real-time on the same timeline every team member sees.
- Git-backed filesystem: All data is stored as markdown files with full version history.
- Self-hosted: Run with Docker Compose; no vendor lock-in.
Quickstart
Prerequisites: Docker, Docker Compose, and an Anthropic (or OpenAI) API key.
git clone https://github.com/kanwas-ai/kanwas.git
cd kanwas
cp .env.example .env
cp backend/.env.example backend/.env
cp yjs-server/.env.example yjs-server/.env
cp frontend/.env.example frontend/.env
docker-compose --profile app upOpen http://localhost:5173 to start working.
Who it's for
- Founders: Combine fundraising decks, customer interviews, MVP specs, and hiring plans into one canvas with agent assistance.
- Product managers: Drop interview snippets, tickets, and competitor screenshots; get a discovery readout and PRD with traceable claims.
- Developers: Pull PM specs and designs onto a canvas; generate an implementation plan with tasks and acceptance criteria, then push markdown to your repo for Claude Code or Codex.
- Marketers: Plan launches with positioning, messaging, and timelines; draft copy variants side-by-side.
- Sales: Maintain reusable account boards with research, comms history, stakeholder maps, and proposal drafts.
Architecture and stack
Built with TypeScript (98.8%), CSS (0.8%), and Docker. Key dependencies: Yjs (CRDT for real-time collaboration), BlockNote (editor), AdonisJS (backend framework), and E2B (sandboxed code execution). The project uses a monorepo with pnpm workspaces.
Community
Join the Kanwas Kollective on Slack for questions and ideas. Contributors should read docs/SYSTEM_OVERVIEW.md first and open an issue before large changes.
📖 Read the full source: HN AI Agents
👀 See Also
Reddit user shares detailed prompt for exporting personal knowledge from AI assistants
A Reddit user has created a comprehensive prompt for extracting structured personal knowledge from AI assistants like Claude, addressing perceived limitations in Anthropic's ChatGPT import feature. The prompt generates three distinct JSON artifacts covering personal knowledge bases, intellectual frameworks, and knowledge graphs.
Zora: Offline-First AI Agent with Default-Deny Security and Local Memory
Zora is an AI agent that runs fully offline via Ollama by default, starts with zero access permissions, and maintains persistent memory across sessions. It addresses security and cost issues seen in other agents.
Vellium Adds Desktop Pets and CLI-Inspired Agents for Local LLMs
Vellium, an open-source crossplatform app for local LLMs, now supports desktop pets that float above windows and agents with MCP integration, terminal commands, and file editing.
ClawProxy: Self-Hosted AI Routing Proxy with Dashboard
ClawProxy is an open-source, self-hosted proxy that centralizes management of multiple AI API keys and models. It provides a unified endpoint, smart key rotation, provider fallback, and real-time logging via a React dashboard.